The Retro Polls Results of polls Correct answers and citations related to each poll Statement of the Retro Poll mission Retro Poll Advisory Board
Retro Poll home


May 2005

The Web of Circumstances
in the 2004 Election1

by Marc Sapir of Retro Poll


On November 2, 2004 most of us participated in a presidential election that drew huge attention in the U.S. and around the world.  Many observers buoyantly noted the success of energized campaigns to register new voters.  American democracy, it appeared, was making a bold comeback from the recent decades of public demoralization, disdain, and disempowerment reflected in low voter registration and turnouts.  Most everyone waited intently to see how many new voters would actually make it to the polls and whether the rightist ideologues in power might be rebuked.  The final vote count suggested that more people voted than in any election since 1968.  Meanwhile, in Redwood City, California, a jury was about to convict one Scott Peterson of murdering his wife.  There would be no smoking gun--in fact no gun at all.  Peterson would be convicted chiefly because of his incriminating behavior with a consort, when no one was looking.  He was being recorded.

In the Presidential election, most of the newly registered voters were registered by liberal organizations-some affiliated with the Democratic Party, but many not.  Many observers, including myself, believed that the higher the voter turnout the greater the likelihood of a Kerry victory.  No one that I know of, however, predicted that almost 17 million more votes would be cast in 2004 than in 2000.  Although several pre-election polls showed W. Bush ahead, this "lead" in the polls was generally proportional to how stringently the polls defined "likely voters".  In polls using a broad definition of "likely voters", crediting the pro-Democrat voter registration effort, the election actually was predicted to be about even, or Kerry slightly ahead.  Given this background, 16.7 million new voters should have given Mr. Kerry an election victory.  The polls by Gallup and others showing Mr. Bush with a significant 7-12 percent lead after the Republican convention were based,  more than "bounce", on a narrow definition of  "likely voter"-presuming very high Republican turnouts and continuing poor Democrat turnouts.   In addition, this election ultimately conformed with the precedent that people who make up their minds in the last week of an election tilt away from the incumbent; most undecideds did vote for Kerry.  Most new voters did also. 

Kerry would also be favored by a strong turnout because most people in the U.S. live in and around large cities.  These cities are the base of the Democratic Party's presidential vote.  The more the full electorate was energized, the higher the proportion of urban voters that should be activated--and the more votes for Kerry--because it is the urban lower middle and working class that has historically the lowest voter turnout record of any segment of the population.  Thus they are the segment with the greatest potential gains in voters if the vote increase were large; indeed, they actually did give Kerry his major support.  Most people who earn under $50,000 per year voted for Kerry.

Yet, somehow, wherever one looked, the larger the vote total in a State or area the more successful was Bush. This appeared to provide increasing margins for George Bush nationwide.  Bush gained 11.5 million total votes over his 2000 showing; Kerry gained 8 million over Gore's showing.2  Of particular interest are Bush's gains in those urban areas where Democratic percent margins did not hold up to 2000 levels.  In large urban counties of a million or more Bush improved by 2.4 percent narrowing his margin of loss to about 55 percent for Kerry to 44 percent for Bush.   In fact an analysis of where Bush gained most over his 2000 support, by Ruy Teixeira of the Century Foundation, (  revealed that, unexpectedly, 19 percent of his improvement came in these large cities with 13 percent of his gain from suburban counties as expected, 26 percent from medium sized cities, 8  percent from small cities, 15 percent from semi-rural counties and 18 percent from totally rural counties.  Now on its face that may appear a "so what?" kind of event.  But improving more in the cities than in suburbia, semi-rural or rural counties is not easy to explain.  

After the polling places closed the major media members of the National Election Pool (ABC, AP, CBS, CNN, Fox) posted the weighted (adjusted for sex, age, ethnicity, and geography) exit poll projections.  These projections had not yet been final adjusted to conform to the local registrars' vote count totals (a process called forcing the numbers) because these were not yet available.  As we now know, these weighted estimates-the best polling sample projections--showed Kerry winning the election by a substantial margin.  When all the counting was done the exit poll numbers were different from the vote tallies by more than one standard error toward Kerry in 26 states and beyond one standard error for Bush in but 4 states. 

Later on, after the exit polls were adjusted to conform to the reported precinct vote totals, Ruy Teixeira reported that the final adjusted exit polls showed Bush making even bigger inroads in the urban centers.  In fact, the disparity he cites between the adjusted exit polls and the vote totals on this geographical-vote-distribution factor is huge.  According to Teixeira, final exit polls show Bush with 13 percent urban improvement over 2000 instead of the 2.4 percent he actually achieved according to vote totals.  The exit polls therefore, not only appear to have been wrong in many of their predictions of the state-by-state victor, but may be skewed in the opposite direction (after final adjustment), with respect to election geography.   Warren Mitofsky of Mitofsky International, which ran the National Election Pool System's exit polls along with Edison Media Research, insists that Teixeira's analysis is invalid because Teixeira compares vote percentages for entire urban counties with the exit polls' single precinct outcomes. but he admits he hasn't looked at whether this actually skews the assessment.  This strange discrepancy suggests--though  doesn't prove--that when exit polls changed their outcomes to conform to the reported precinct vote totals at 1:40 a.m. on November 3, this adjustment forced the exit poll urban vote totals to appear more favorable to Bush than was actually the case. That is hard to explain unless the official precinct based vote totals are not real and the adjustment pushed some of the parameters out of sync. 

The exit poll data analysis by Edison Media Research and Mitofsky International3 was based upon voter surveys taken at 1,250 precincts nationally4 and included over 75,500 voters responding to state and national questionnaires.  Of these 1,250 precincts 341 (27%) were in cities over 50,000.  As of this writing, the detailed precinct specific data breakdowns have just been posted on line by the Inter-University Consortium for Political and Social Research based at the University of Michigan (   Will in the months ahead further data analysis be able to clarify whether certain urban voter groups did or did not shift to Bush at much higher rates than other population groups and at higher rates than projected in pre-election polls?  Probably not, because the data is restricted by "confidentiality" so that it is not possible to ascertain which Exit Poll precincts correspond to which real precincts and the raw data lacks information about the weights that were used. A specific exit poll finding (on Latinos) will be discussed below in the section: Big Cities and Centralized Computer Vote Tabulation.

Exit Polls: Are They Accurate?

Many election observers and statisticians believe that exit polls are more accurate than conventional telephone survey opinion polls at predicting vote outcomes.  The voters are real voters not "likely voters"; a relatively high percent of voters agree to participate; and "which candidate you voted for?" isn't a question subject to bias in the wording or placement of the question.  The accuracy of exit polls has probably been overstated, however, because these polls use "cluster" sampling in which, first, precincts are selected at random, and then voters, thus adding an extra layer of potential sampling error.  Nevertheless, exit polls are historically relatively accurate in predicting outcomes, tending to fail frequently only when candidate totals are close.5  Shockingly, in this election, 10 of 11 State exit polls in the "battleground" states predicted Kerry with a stronger vote than the final results. Though two of those State polls were not way off, the three key states of Pennsylvania, Ohio and Florida were off way beyond statistical significance (as demonstrated in a very precise paper by Jonathan Simon, J.D. of Verified Vote and Ron Baiman PhD at the Institute of government and Public Affairs, University of Illinois at Chicago-see their posting at  In general, the exit polls in the battleground states were stronger for Kerry compared with those in other states.  In Florida, for example, the exit poll showed Kerry winning the State by  5 percent, though Bush won by almost 3 percent.  In Pennsylvania the exit poll discrepancy was 8.4 percent and in Ohio 10.6. percent on the spread between the two candidates.  In addition, for three selected states that relied extensively on paper ballots (Illinois, Maine, Wisconsin) the exit polls corresponded closely to the final tally as compared with battleground states that relied extensively on electronic touch-screens, where the vote tallies invariably favored Bush (Lewis Lapham, January edition of Harper's pp. 10 -11)..

Steven Freeman, a professor of innovation and research methods at University of Pennsylvania, originally  suggested that exit polls are so accurate that the chances of all these battleground states outcomes being consistently wrong  on one side in 10 of 11 cases are one in 250 million (   John Allen Paulos, Professor of Mathematics at Temple University, wrote that the probability of the percent discrepancies occurring simultaneously in just Ohio, Pennsylvania and Florida is about 1 in a million (Philadelphia Inquirer, 11/24/04).  A number of mainstream polling people and statisticians critiqued Freeman's paper.  They focused on the issue of non-randomness of individuals based upon the precinct samples. Freeman later took their concerns into consideration and he revised the possibility of the results occurring by chance down to 1 in 650,000.  Regardless, Freeman's detractors (as Simon and Baiman point out at present no compelling evidence (they presented instead conjecture and hypotheses such as Republican distrust of pollsters and Democrats voting earlier in the day etc). to explain why all these different state exit polls had Kerry always running stronger than the outcome.  Edison/Mitofsky's analysis concludes that younger interviewers were less effective at getting completed surveys and their data was more shifted toward Kerry.  But their paper shows that even interviewers over the age of 35 had an average within precinct error over 5.5 percent toward Kerry--hardly a convincing explanation.  Since Freeman's and Poulos' estimates can not pinpoint the source of a systematic error or bias we still lack any compelling explanation of why all the deviations go in the same direction.  Nor have the critics of Freeman explained why the exit polls results were more consistent in non-battleground and non-electronic machine dominant states.

More Pieces of the Same Puzzle

When Richard Hayes Phillips a PhD geomorphologist, analyzed ( ), the actual precinct vote totals in and around Cleveland (Cuyahoga County) he found puzzling results as well. Cleveland is a relatively liberal community.  It's Dennis Kucinich's base. Cleveland has a large African American population and a number of prominent Universities.  Kerry won Cuyahoga county but his margin was reduced from that of Al Gore and reduced from that expected.  Furthermore, the turnout in many of the precincts where Kerry won were remarkably low-as low as 7.3 percent and 12 percent.  Hayes Phillips already knew about the protests lodged over long lines in some of these areas where people were still on line 4 hours after the polls closed, and many frustrated voters just walked away.  On the other hand, turnout in precincts that went for Bush in this county were up over 70 percent.  Overall turnout in precincts supporting Bush was 20 percent higher than turnout in precincts supporting Kerry.   This was in spite of massive and successful registration and "get out the vote" (GOTV) efforts by Kerry supporters centered on Cleveland. 

Hayes Phillips then compared the turnout per precinct to the voter density per election machine allocated to the precincts.  Lo and behold there was a strong and clear relationship between voter turnout and number of voters per machine.  The more voters per machine the lower the actual voter turnout.  He calculated that in most Republican areas one would have 7 minutes, on average, of machine-time per voter (if the flow was constant all day).  If the vote was sporadic you'd have less time. In many of the Democrat strongholds a doubling of the voters per machine would force long lines and major waits (which went up to 6 and 8 hours).   Analysis of the voting patterns in the capital, Columbus, Ohio, revealed the same thing.  The Republican registrar in Columbus had even made public statements that he was going to move machines out of the more heavily Democratic areas, thus creating the inequality confirmed by Hayes Phillips analysis.

In England, Elizabeth Liddle, a PhD candidate with statistical experience, took Hayes Phillips findings and ran "regression analysis" on the data to see if she could prove that the effect Hayes Phillips revealed was statistically significant.  Liddle's work showed more than that.  As Hayes Phillips predicted, the correlation of voter turnout percentage in a precinct is strongly tied to the number of registered voters per machine in that precinct.  If you had too few machines the final voter turnout was lower. And despite some overlap the number of voters per machine is strongly tied to who was going to (and did) win the precinct.  The result is a net loss of thousands of votes for John Kerry.  Liddle made a clever additional adjustment to be sure her calculations were extremely conservative.  She re-analyzed the data assuming that Democratic turnout would be lower than Republican turnout based upon past performance.  The disparity showing disenfranchisement of Kerry voters remained and was statistically likely to occur by chance less than one time in a million.  Interestingly she writes that both Bush and Kerry lost votes as a result of inadequate machine to voter ratios in Franklin County, but in her most conservative model Kerry lost 7,000 votes more than Bush in Franklin County alone (where Columbus is located).  Even given this apparently successful effort at vote suppression in pro-Kerry precincts, voter turnout numbers like the 7 percent, 12 percent, 18 percent, 21 percent and 23 percent in precincts back in Cuyahoga County are hard to imagine in this particular election with its large turnout.  It suggests other methods of vote suppression and that requires independent investigation (not by a partisan Republican Secretary of State).   One  Cuyahoga example has more recently come to light.  The Republican's vote-challenging and legal restriction efforts worked. Over 25,000 voters in Cuyahoga county were forced to use Provisional Ballots.  About a third of these were disqualified and of 2,100 that were alleged to have voted out of their precinct (a practice allowed in past elections) case-by-case checking has showed that 1,200 of these were actually in their correct polling place and thus had their votes illegally disqualified (

Thus, in at least two major urban areas of Ohio the intentional allocation of voting machines to increase voter density in strong Kerry areas can explain George Bush's improved percentages in the Democratic controlled urban centers.  The fact of discrimination against certain classes of voters, most of whom went to the polls to vote for John Kerry can not be ignored.  On December 15, a month into major protests and investigation of these irregularities, the New York Times and Washington Post finally noticed that there were concerns in Ohio about the fairness of the election, but they presented these as relatively anecdotal and non-determinative issues, eventually burying the story.  At the same time, Ohio's county and state registrars began to promote the idea that machine allocation problems arose because they based their assignments upon past performance.  In fact, this argument favors fraud because there were not lines at most of the Republican precincts in most of the counties.  If voting was way up over expected only in the heavily Democratic precincts with the long lines how did Mr. Bush expand his support so successfully?  Remember, a large Bush voter expansion over 2000 was necessary for him to win and should have created the same long lines.  The explanation provided by the Republican registrars suggests a cover up.   Other investigations by the Columbus Free Press have turned up several strongly pro-Bush precincts in Ohio where Bush took over 80 percent of the votes but the vote turnout was as high as 98.5 percent.   After calling several hundred registered voters in one precinct to inquire if they voted the Free Press found enough non-voting people to prove the turnout rate, and thus some percent of the large Bush vote in those precincts, was fabricated (  

Big Cities and Centralized Computer Vote Tabulation

At University of California, Berkeley, students of Professor Mike Hout, an award winning sociologist decided to analyze Florida voting patterns in this election in comparison to 2000 (   They discovered that Bush made surprising percentage gains in Florida's major cities, similar to those in Cleveland and Columbus and similar to the reported national average; but unlike Ohio, Florida's large cities used the no-audit electronic voting machines.  Kerry's urban support seemed to flag unexpectedly with decreased victory margins in the usual Democratic strongholds of Palm Beach and Broward and Miami Dade where he barely won.  How do we know what was expected in 2004?  One measure of expectation is past performance, another is current performance in general.  Hout et al's factor of interest to us is the type of voting machine (the electronic machines without audit trail).  They found that the change in percent of votes for Bush in each county from 2000 to 2004 differed by voting machine type.  In comparison with how Bush did throughout Florida (in the non-touch-screen counties) his performance in these major touch-screen counties was significantly better.  Bush seemed to have about 130,000 votes more than he should have in these urban touch-screen counties.  If the votes gained by Bush were actually votes that were cast for Kerry, the Hout group calculated that about 260,000 votes of the 380,000 vote Bush margin in Florida would disappear.  

The UC group was concerned that maybe some other innocent factor about the touch-screen counties was really behind this shift.  So they built up a mathematical model to include possible alternative causes, from county size to ethnic percentages to past election performance.  When those factors were included in their model, the machine type still came up as a separate independent factor associated with the Bush gains.  As time passed, various adaptations of their model did not change the effect of machine type  as a statistically significant predictor of Bush's surge.  And where were these machines?  In the major population centers.  Some academicians have raised objections to this work, but Hout et al have responded effectively to each concern. ( ).

Yet even before the Hout group went public, another group was working on a different theory concerning the Florida results.  They reported the unexpected finding of heavy voting for Bush in many counties with large Democratic voter registration.  At first some critics attributed this to the effect of local custom in small rural counties from the Dixiecrat days.  Conservative rural folks still remain in the Democratic Party for local election purposes, but are really Republicans and Bush supporters. However, Elizabeth Liddle found that those rural counties were not necessary to show an unexplained Democrat shift to Bush throughout Florida ( .  The shift of Democratic registrants to Bush was also seen in medium sized and other urban counties as well. 

The Phenomenon of "Pseudo-crossover voting" in Republican Strongholds

Researcher Oliver Dawshed went on to describe two other findings: 1) as in 2000 with Gore there were  particularly high ballot spoilage rates (undervotes--ballots discarded and not counted) only in some strongly pro Kerry areas, and 2) an unexplainable "pseudo" cross-over of major proportion of Democrats who voted for the Democratic candidate for U.S. Senator (Betty Castor) but not for Kerry in counties that use optical scanning equipment (which do have an audit trail but are tabulated centrally by computer just like the electronic machines) (  Dawshed noted that Betty Castor, who barely lost her bid to become a Democratic U.S. Senator from Florida, did remarkably better than Kerry in many of these counties.  Perhaps Castor simply was viewed as more moderate than Kerry?   But, in fact, a broad smear campaign was carried out to label Castor as a liberal pro-abortionist who supported gay marriage.  If the Bush base turned out in droves it surely wanted to defeat her as well.  Based upon the cross-over totals, tens of thousands of votes may have been switched from Kerry to Bush in the optical scan tabulation process.  And since Hout and his students suggested that the touch-screen counties had significantly more votes for Bush than predicted these two different factors could potentially interact in a way that would be more than additive.  The number of votes predicted to have been switched by each model could be an overestimate or an underestimate, but if these Florida findings are correct they might account for the 380,000 Bush margin. The exit polls showed Kerry winning Florida by 5 percent.  If that figure were to be accurate, it would have required major vote switching.  The corporate media and even many Democratic politicians and pundits keep saying:   "let's look into these things for next time, but the election's outcome does not seem to be in doubt."  This statistical work suggests otherwise. 

Meanwhile a revelation similar to the Dawshed "pseudo" crossover was reported in Ohio and excited Jesse Jackson's concern.  A Black woman judge from Cleveland (C. Ellen Connally) running for Chief Justice of the Ohio Supreme Court on the very bottom of the ballot outpolled John Kerry throughout southern Ohio, by tens of thousands of votes in some counties.  Typically, vote tallies diminish as one goes from the most prominent contests such as the Presidency on down the ballot to the least well known or understood candidates.   C. Ellen Connally is not  a Republican.  She is shown on her web site with union leaders and attending a protest of an appearance by Dick Cheney in Cleveland.  The southern region of Ohio includes heavily Republican counties where Bush built up (as expected) much of the cushion he needed to overcome Kerry's advantage in the major cities.  But of course the state's voting process is totally in the hands of Republican party operatives under the direction of  J. Kenneth Blackwell, Ohio's Republican Secretary of State, the well known  State co-chair of the Bush campaign in Ohio.  Blackwell is the man who, while stalling a month to certify the Ohio election, successfully blocked the vote investigation by lawyers for the Green and Libertarian party candidates and ignored a subpoena requiring his testimony before the Ohio Supreme Court.  Blackwell may yet offer a brilliant explanation as to how it might have occurred to so many registered Democrats in Southern Ohio to vote for Connally but not for John Kerry.  But as with Castor, Connally was generally considered a liberal in this race and should not have outpolled Kerry.     

The Pivotal Urban Vote and the Machinery of the Election

There is no doubt that the above reports are merely forms of circumstantial rather than direct evidence.  In this election over 200,000 calls were logged to a major organization monitoring problems and improprieties; 38,061 documented incidents were reported and can be viewed at vote protect (  and of the hundreds that have been verified, almost all resulted in vote reduction for Kerry or vote inflation for Bush.  There is still not conclusive proof of a broad conspiracy to steal the Presidential election--although there  is proof of illegal activity before and during the election by Republican operatives.  But there is a compelling thread of circumstance focusing on the urban vote.   

Despite the possibility of vote switching in non-urban areas as well, it's important to focus on the incongruity of major Republican increases in vote totals in areas where the Democrats have their base and were highly active in 2004.  This incongruity has to be viewed skeptically and thoroughly investigated. 

Of course, this urban vote problem is merely a hypothesis.   If George W. Bush somehow made unique inroads into traditionally Democratic strongholds in major cities that might explain a good many of the anomalous findings.  Let's look at demographics to try and confirm or debunk such a hypothesis.  According to outcome and exit poll data there were four demographic groups among which Bush made surprising gains.  They were Catholics, white women, Latinos, and the elderly.  A look at the Latino projection is particularly informative.  In 2000, the Annenberg (pre-)Election Survey showed 35 percent voting for Bush.  This time they found 41 percent.  But what percentage of Latinos actually voted for George W. Bush.  Below is a chart of widely discrepant exit polls in the 2004 election.

Surveys of Bush support among "Hispanics"2004

William C. Velasquez Exit Poll

31.4 %

LA Times National Exit Poll


National Annenberg Election Survey


The National Election Pool (NEP) National Survey


The NEP National + State Cross Survey Analysis


The Edison/ Mitofsky explanation for these wide ranging discrepancies pertains to varying definitions of who is included as "Hispanic" and is too detailed to elaborate here.  However, one key fact their evaluation highlights has direct bearing on the urban vote issue.  To debunk the validity of the Velasquez finding of weak support for Bush, Edison/Mitofsky note that Velasquez chose 56 precincts with high Latino voter roles in only 14 states, presumably mostly urban precincts.  In comparison, they point out, the NEP National Exit Poll shows Bush with only 29% Hispanic support in the urban centers of population greater than 500,000 (even though the overall survey gave him 42%).  But what this analysis actually shows us is that, as with African Americans whose 2004 shift to Bush was a statistically insignificant 1-2 percent (91 percent to 89% in the NEP exit polls), Latinos apparently did not contribute significantly to the big urban shifts that gave Bush the election.  That leaves only three other urban groups to counteract the 89 percent African American, the 71 percent Latino and the majority working class vote for Kerry.   We could determine if Bush made major gains among these groups if we had voting machines with an audit trail in such counties.  Unfortunately, as in South Florida's major urban centers this is often impossible because those in power chose to make it impossible to audit.  In Florida in particular that decision by Governor Bush violated a State court order and was in defiance of the rules of Florida's new recount law. 

On page 39 of the Edison/Mitofsky January 19 NEP exit poll evaluation report one finds that "some have suggested that exit poll data could be used as evidence of voter fraud by showing error rates were higher in precincts with touch screen and optical scan voting equipment...In our exit poll sample overall, precincts with touch screen and optical scan voting have essentially the same error rates as those with punch card systems.  In the larger urban area these systems had lower within precinct error than punch card precincts."  To begin with, an incisive February statistical critique of the Edison/Mitofsky report by a group of established academic statisticians and mathematicians ( points out that except for paper ballots all other methods including the punch cards and optical scanners are usually being centrally computer tabulated, and so could be equally tampered with.  However, and leaving that issue aside, Edison/ Mitofsky present a chart on the following page intended to graphically demonstrate that in large urban areas touch screen precincts were less anomalous than optical scanning precincts.  However, the chart does not support that quote.  Instead one finds that in a comparison of voting method and exit poll error margin between urban and rural areas, the suburbs-a Bush stronghold-are buried within the urban category.  Not only that, but of the 827 precincts analyzed in the urban category about 60% of them were suburban.  To present this chart as evidence against fraud or to support any argument is certainly inappropriate.  You will recall that only 13% of Bush's overall gains came from the suburbs.  Yet the suburbs with 487 precints included is far and away the largest subsample in the exit poll.  As the uscountvotes paper suggests it is certainly easy to assert that there is no evidence of fraud in the 2004 election if one is intent upon avoiding the opportunity to look for it.  I had hoped that with the exit poll case data now available to anyone with the SPSS computer program, analysts could compare the demographic voter breakdown in the large and medium sized cities and to look for other potential discrepancies in the data.  But, as mentioned earlier, enough information has been withheld to block, for now, that effort. 

Beyond the exit polls, across the country there are many large cities.  The Hout model can be run comparing those urban centers that used touch screen no-audit machines and those that used other centralized and  decentralized voting tabulation methods.   If this is done it will either prove that touch screen non-auditable voting was consistently a factor related to Bush's capturing the national popular vote throughout the country or it will show non-uniformity. (i.e. no consistent relationship to these machines around the country).  If the model fails to show consistency  it would not rule out selective tampering.  However, if Bush won legitimately, we should see some large cities using manual balloting or audited decentralized tabulation which verify that Bush still captured more than expected votes in those major cities and among those demographic groups as well  Those cities'/counties' ballots are easy to audit to determine if the demographic outcome data fits with what we have been told--a determinative switch of Catholics, white women, and the elderly in large cities to George W. Bush.  Will either dominant political party willingly support a limited investigation of this kind that can explain a good deal of the aberrant and anomalous voting pattern and clear many of the allegations of fraud?  Don't hold your breath.

Argument: The Case for Theft-Rove's Playbook

The White House's Karl Rove believes that the best defense against an adversary is to attack, applying the Powell military doctrine of overwhelming force to the art of politics.   Rove preaches that you attack your adversary by undermining his strengths.  Thus, the key battle to be won in attempting to derail John Kerry's campaign was to attack his Vietnam military service while distracting from Bush's behavior at the time.  Out came the Swift Boat veterans who know that John Kerry betrayed them and his country because he then spoke out against the war in Vietnam. 

On election night, Karl Rove declared to the gathered faithful in the White House that George Bush had won Ohio-and the presidency--before there was any way for him to know the actual results (unless the Ohio election was rigged).  A cheer went up and then this "fact" was announced to the media.  This announcement was made not that long after J Kenneth Blackwell said that at least 250,000 provisional and absentee ballots remained uncounted with Bush ahead by 136,000 and, shortly thereafter, tried to retract that statement and change the number to 150,000, where it remained permanently.  (Maybe Rove got nervous and preempted).

But is Karl Rove's declaration of victory evidence of anything, really?  Karl Rove's general behavior is a known-a biography and movie document it. Rove said, when he first heard that exit polls projected Kerry the big winner, that it made him very "angry"--not concerned, worried, disappointed or anxious but "angry" (as if the plans he had laid had not been well implemented).   There was vicious and illegal intervention in the election-e.g. Black voters were notified to vote on the wrong days; were warned that they would be arrested at the polls if they had outstanding tickets, warrants or child support payments; were told that someone would collect their vote for them.   Why would one put in place lies and illegal activity to win an election without a strategic plan to switch votes if necessary?   Rove aimed to assure Bush an absolute national majority as well as an electoral vote majority.  And that would require weakening the vote for Kerry at his very strength, where most of the people in the U.S. live and where Kerry had to gain overwhelming superiority to become President, the major cities.  This could be done partly by "false advertising" but might also require vote tampering.

What I call the urban strategy may explain why, on that November 2 evening, Republican PR operatives explicitly declared  that they saw big things happening in the rural areas (such as the Florida panhandle) that would push Bush over the top despite exit poll projections to the contrary.  In highlighting the expected winning Bush totals in rural areas which were not particularly exciting, they distracted attention from the fact that the election was won by crippling Kerry in the big cities, just as they had crippled him with the veterans' vote while destroying the real Vietnam story-namely, Bush had refused to take a physical in the National Guard, had been grounded though he was needed as a pilot, and had not only avoided Vietnam service but disappeared.   And because the Republicans have been allowed by the media and the Democrats to act with such impunity both in 2000 and 2002 in Florida they knew they could get away with it again.  

In summary, it's a circumstantial case:  Because he apparently only won about 7 of the 16.7 million new voters, George W. Bush had to find another 4.5 million votes from those who voted for Al Gore and Ralph Nader in 2000.   Vote totals and the exit polls suggest he found many of them in large and medium sized cities.  But did he?  1) There's the exit polls showing 10 of 11 "battleground" states significantly more favorable toward Kerry than the outcomes, and the actual vote totals also showing Bush with more inner city votes than a Republican (barring a landslide) ought to have; 2) the Hout analysis of Bush over-votes in counties with electronic voting machines possibly accounting for switching of 130,000 Kerry votes to Bush; 3) the huge overall increase in Bush votes relative to the Democrats compared with 2000 even though he did not win a majority of new voters; 4) the "pseudo" crossover votes of Democrats voting for Bush by the tens of thousands in Ohio and Florida; 5) the high ballot spoilage rates only in Kerry dominant precincts and counties; 6) the vote suppression by the systematic reduction of voting machines in many pro-Kerry precincts in Ohio; 7) the Bush-heavy precincts in Ohio with more votes than voters, yet no similar pro-Kerry precincts; and 8) the foot dragging and obstruction by J. Kenneth Blackwell reminding us of the infamous behavior of Ms. Katherine Harris in Florida in 2000 that covered up the disenfranchisement of tens of thousands of African Americans.   These various findings suggest not only systematic disenfranchisement of many thousands of voters, but the fraudulent misappropriation of actual votes.  It may be years before we can definitively determine if/that tabulation software was tampered with across the nation, yet the fact that the two companies responsible for the non-auditable machines and software are controlled by the Bush team is a travesty.  I have proposed a simple audit maneuver that can be used to look at the national vote and see if shifts in the votes of Catholics, women, and the elderly are real or not. 

Beyond the local level, fair and democratic elections in the U.S. may already be all but impossible, even without fraud.  The vast sums of money and media support required to engage in a national campaign are only available to those willing to do the bidding of moneyed interests.  The exclusion of third parties in major debates, the closed nature of the two party system, the anti-populist biases of the major corporations that own the media outlets, the absence of a direct presidential vote, all mitigate against the democratic inclination of our people.  The issues in the campaigns are at best caricatures of real the issues. Yet, we can see from the 2004 election that the public is both concerned and more polarized than at any time since the McCarthy era.  The very fact that open voter suppression has returned to the U.S. with the 2000, 2002 and 2004 elections is evidence that the people are trying to be heard from and taken seriously.  It's the responsibility of everyone of us to insist upon the right of all to participate in determining the future.  The right to vote is as good a place to start as any. 
© May 2005 by Marc Sapir

Note: Special thanks to Warren Gold, James Murphy, and Robert Godfrey for help in the preparation of this article.

Marc Sapir is a family practice physician in Berkeley.  He directs Retro Poll ( and has recently published a novel, the Last Tale of Mendel Abbe: Sonny Bush and the Wise Men of New Chelm.


1 Readers will note that this article avoids specific discussion of anecdotal events that have been reported such as two election machines that were found to have given George Bush several thousand over-votes and a few reports by Kerry voters that machines tabulated their votes for Bush.  While such incidents might contribute to the "Web of Circumstance" they do not directly pertain to the core thesis of the article and so would add little in this context.
2 A most puzzling finding is discrepancy between AP-Washington Post final vote totals versus those of David Leip's election atlas and CNN's web site.  Leip posts almost 4 million more total votes.  Bush's improvement from 2000 to 2004 thus ranges from 8.5-11.5 million votes and Kerry's improvement from Gore ranges from 4.5-8 million votes depending upon which data are used.  David Leip's data comes directly from the vote totals reported on official State Election web sites. 
3 Evaluation of Edison/Mitofsky Election System 2004 prepared by Edison Media Research and Mitofsky International for the National Election Pool (NEP), January 19, 2005.
4 An absentee voter sample from these precincts was also contacted by other means.  
5 Three other  Presidential  exit poll in the past 20 years were off by about two percent with one (1992) off by 5 percentage points, still 25% less than the current polls' average deviation. 


Retro Polls The Retro Polls Results of polls Results of polls Correct answers and citations related to each poll Correct answers and citations related to each poll Statement of the Retro Poll mission Statement of the Retro Poll mission Retro Poll Advisory Board Retro Poll Advisory Board Retro Poll home Retro Poll home